Skip to main content

AI Architecture

Access

The AI Architecture tab is visible to Admin, Owner, and Advisor roles only. Members and Support agents do not see this tab.

The AI Architecture canvas gives you a visual, bird's-eye map of your AI Agent's complete setup. Every component your agent uses is shown as a connected node: the LLM provider, memory layer, actions, knowledge base topics, and human handoff. You can explore, navigate, and even test the agent directly from this view.

The AI Architecture canvas gives a complete picture of your agent's setup at a glance.

Canvas nodes

Each node on the canvas represents a part of your agent's configuration. Clicking most nodes navigates you directly to that feature's settings page.

LLM Provider

Displays the current AI model provider and model your agent uses (for example, OpenAI GPT-4o or Anthropic Claude). Click it to jump to Basic Info settings and change the model.

Memory

Shows whether agent memory is enabled. When active, the node is highlighted; when disabled, it appears muted. Click it to go to Agent Memory settings.

AI Agent (center)

The central node representing your agent. It shows:

  • Agent name and icon
  • Active / Inactive toggle — hover over the agent node to reveal the toggle. Slide it to activate or deactivate the agent. A confirmation dialog appears before any status change takes effect.

Actions

An expandable hub node. Click it to expand and reveal the action type categories your agent supports:

Action typeDescription
WorkflowBoltic/Vortex workflows connected to this agent
Boltic MCPMCP actions powered by Boltic's built-in provider
Custom MCPCustom MCP servers you have configured
MarketplaceBoltic integrations from the marketplace
SkillReusable skills attached to this agent (shows a Tools list)
AgentSub-agents this agent can delegate to

Click any action type node to expand it and see individual actions. Hover over an action card to reveal the delete button (trash icon). Clicking delete opens a confirmation dialog that requires you to type the exact name of the action before deletion is allowed — this prevents accidental removal. An Add new node with a dashed border appears at the end of each list so you can create a new action of that type without leaving the canvas.

Click the Actions node to expand it, then click a type to see individual actions.

Knowledge Base

An expandable hub node for your agent's knowledge topics. Click it to see the topics attached to this agent (up to 5). Hover over a topic card to reveal the delete button, and an Add new node with a dashed border lets you create a new topic directly.

Expand the Knowledge Base node to view and manage knowledge topics.

Human Handoff

Shows whether human handoff (customer support escalation) is enabled. When active, the node is shown in green. Click it to go to Support settings and configure handoff rules.


Canvas controls

A control bar is fixed at the bottom center of the canvas:

ControlWhat it does
⇄ / ⇅Toggle between horizontal (left-to-right) and vertical (top-to-bottom) layout
Zoom out
% displayShows the current zoom level
+Zoom in
Fit the entire canvas to the visible screen area

You can also drag any empty area to pan the canvas, and use scroll or pinch gestures to zoom.

Use the bottom controls to zoom, pan, or switch between vertical and horizontal layout.
Layout preference is remembered

Your chosen layout direction and which nodes are expanded are saved per agent for the duration of your browser session. They are restored the next time you open the AI Architecture tab.


Test Agent playground

Click the Test agent button in the top-right corner to open an inline chat panel on the right side of the canvas. This lets you send messages to your agent and see its responses without navigating away.

The Test Agent panel slides in from the right so you can chat with your agent while viewing its architecture.

Click Close agent to dismiss the panel and return to the full-width canvas.

note

The test playground uses the same agent token and configuration as the live deployment. Responses reflect the agent's current knowledge base, actions, and LLM settings.


Every node is a shortcut to the relevant settings. Instead of using the sidebar, click a component on the canvas to jump directly to its configuration:

NodeNavigates to
LLM ProviderBasic Info → LLM Settings
MemoryAgent Memory settings
AI AgentBasic Info settings
Actions (hub)Expands in-place
Individual actionActions page
Add new actionAction creation flow for that type
Knowledge Base (hub)Expands in-place
Individual topicTopic datasources page
Add new topicTopic creation
Human HandoffSupport settings

Desktop only

The AI Architecture canvas is not available on mobile screens. Use a desktop or laptop browser to access this feature.

Kaily logo

More than just a virtual AI assistant, Kaily brings interactive, human-like conversations to your website. Easy to create, easier to customize, and the easiest to deploy—no code required. Let Kaily enhance your user experience using the information you provide.

Is this page useful?